From 082d7e8407b1f73f1b7f3b1a3082e0ccf4783c5a Mon Sep 17 00:00:00 2001
From: Roman Joeres <Roman.Joeres@helmholtz-hzi.de>
Date: Mon, 19 Aug 2024 16:06:44 +0200
Subject: [PATCH] Started full downstream training

---
 configs/downstream/all.yaml | 96 ++++++++++++++++++-------------------
 gifflar/metrics.py          |  2 +
 gifflar/utils.py            |  4 +-
 3 files changed, 52 insertions(+), 50 deletions(-)

diff --git a/configs/downstream/all.yaml b/configs/downstream/all.yaml
index e9255b2..d5bd024 100644
--- a/configs/downstream/all.yaml
+++ b/configs/downstream/all.yaml
@@ -9,30 +9,30 @@ datasets:
     task: classification
   - name: Taxonomy_Domain
     task: multilabel
-  #- name: Taxonomy_Kingdom
-  #  task: multilabel
-  #- name: Taxonomy_Phylum
-  #  task: multilabel
-  #- name: Taxonomy_Class
-  #  task: multilabel
-  #- name: Taxonomy_Order
-  #  task: multilabel
-  #- name: Taxonomy_Family
-  #  task: multilabel
-  #- name: Taxonomy_Genus
-  #  task: multilabel
-  #- name: Taxonomy_Species
-  #  task: multilabel
+  - name: Taxonomy_Kingdom
+    task: multilabel
+  - name: Taxonomy_Phylum
+    task: multilabel
+  - name: Taxonomy_Class
+    task: multilabel
+  - name: Taxonomy_Order
+    task: multilabel
+  - name: Taxonomy_Family
+    task: multilabel
+  - name: Taxonomy_Genus
+    task: multilabel
+  - name: Taxonomy_Species
+    task: multilabel
 pre-transforms:
 model:
-  #- name: rf
-  #  n_estimators: 500
-  #  n_jobs: -1
-  #  random_state: 42
-  #- name: svm
-  #  random_state: 42
-  #- name: xgb
-  #  random_state: 42
+  - name: rf
+    n_estimators: 500
+    n_jobs: -1
+    random_state: 42
+  - name: svm
+    random_state: 42
+  - name: xgb
+    random_state: 42
   - name: mlp
     hidden_dim: 1024
     batch_size: 256
@@ -41,29 +41,29 @@ model:
     patience: 30
     learning_rate: 0
     optimizer: Adam
-  #- name: sweetnet
-  #  hidden_dim: 1024
-  #  batch_size: 512
-  #  num_layers: 16
-  #  epochs: 1
-  #  patience: 30
-  #  learning_rate: 0.001
-  #  optimizer: Adam
-  #  suffix:
-  #- name: gnngly
-  #  hidden_dim: 1024
-  #  batch_size: 512
-  #  num_layers: 8
-  #  epochs: 1
-  #  patience: 30
-  #  learning_rate: 0.001
-  #  optimizer: Adam
-  #  suffix:
-  #- name: gifflar
-  #  hidden_dim: 1024
-  #  batch_size: 256
-  #  num_layers: 8
-  #  epochs: 100
-  #  learning_rate: 0.001
-  #  optimizer: Adam
-  #  suffix:
+  - name: sweetnet
+    hidden_dim: 1024
+    batch_size: 512
+    num_layers: 16
+    epochs: 100
+    patience: 30
+    learning_rate: 0.001
+    optimizer: Adam
+    suffix:
+  - name: gnngly
+    hidden_dim: 1024
+    batch_size: 512
+    num_layers: 8
+    epochs: 100
+    patience: 30
+    learning_rate: 0.001
+    optimizer: Adam
+    suffix:
+  - name: gifflar
+    hidden_dim: 1024
+    batch_size: 256
+    num_layers: 8
+    epochs: 100
+    learning_rate: 0.001
+    optimizer: Adam
+    suffix:
diff --git a/gifflar/metrics.py b/gifflar/metrics.py
index d23945f..b52bbb2 100644
--- a/gifflar/metrics.py
+++ b/gifflar/metrics.py
@@ -10,6 +10,8 @@ def __init__(self, threshold=0.5, dist_sync_on_step=False, **kwargs):
         self.add_state("false_negatives", default=torch.tensor(0), dist_reduce_fx="sum")
 
     def update(self, preds: torch.Tensor, target: torch.Tensor):
+        print(preds.shape)
+        print(target.shape)
         preds = (preds > self.threshold).int()
         target = target.int()
 
diff --git a/gifflar/utils.py b/gifflar/utils.py
index 73987cb..08a6768 100644
--- a/gifflar/utils.py
+++ b/gifflar/utils.py
@@ -15,7 +15,7 @@
 from sklearn.svm import LinearSVR, SVC, SVR
 from torchmetrics import MetricCollection, Accuracy, AUROC, MatthewsCorrCoef, MeanAbsoluteError, MeanSquaredError, \
     R2Score
-from gifflar.metrics import Sensitivity
+from gifflar.sensitivity import Sensitivity
 
 
 # MASK: +1 | Other: +2 => len(...) | len(...) + 1
@@ -116,7 +116,7 @@ def get_metrics(
             Accuracy(**metric_args),
             AUROC(**metric_args),
             MatthewsCorrCoef(**metric_args),
-            # Sensitivity(**metric_args),
+            Sensitivity(**metric_args),
         ])
     return {"train": m.clone(prefix="train/"), "val": m.clone(prefix="val/"), "test": m.clone(prefix="test/")}